fmicb-10-00093 February 12, 2019 Time: 16:37 # 1

ORIGINAL RESEARCHpublished: 13 February 2019

doi: 10.3389/fmicb.2019.00093

Edited by:Baltasar Mayo,

Spanish National Research Council(CSIC), Spain

Reviewed by:Giulia Tabanelli,

University of Bologna, ItalyErica Renes,

Universidad de León, Spain

*Correspondence:Elena Franciosi

elena.franciosi@fmach.it

Specialty section:This article was submitted to

Food Microbiology,a section of the journal

Frontiers in Microbiology

Received: 08 October 2018Accepted: 16 January 2019

Published: 13 February 2019

Citation:Carafa I, Stocco G, Nardin T,

Larcher R, Bittante G, Tuohy K andFranciosi E (2019) Production

of Naturally γ -AminobutyricAcid-Enriched Cheese Using the Dairy

Strains Streptococcus thermophilus84C and Lactobacillus brevis DSM

32386. Front. Microbiol. 10:93.doi: 10.3389/fmicb.2019.00093

Production of Naturallyγ-Aminobutyric Acid-EnrichedCheese Using the Dairy StrainsStreptococcus thermophilus 84C andLactobacillus brevis DSM 32386Ilaria Carafa1, Giorgia Stocco2, Tiziana Nardin3, Roberto Larcher3, Giovanni Bittante2,Kieran Tuohy1 and Elena Franciosi1*

1 Research and Innovation Centre, Food Quality and Nutrition Department, Fondazione Edmund Mach (FEM),San Michele all’Adige, Italy, 2 Department of Agronomy, Food, Natural Resources, Animals and Environment(DAFNAE), University of Padova, Legnaro, Italy, 3 Technology Transfer Centre, Fondazione Edmund Mach (FEM),San Michele all’Adige, Italy

The cheese-derived strains Streptococcus thermophilus 84C isolated from Nostranocheese, and Lactobacillus brevis DSM 32386 isolated from Traditional Mountain Malgacheese have been previously reported as γ-aminobutyric acid (GABA)-producers in vitro.In the present study, the ability of these strains to produce GABA was studied inexperimental raw milk cheeses, with the aim to investigate the effect of the cultureand the ripening time on the GABA concentration. The cultures used consisted onS. thermophilus 84C alone (84C) or in combination with L. brevis DSM 32386 (84C-DSM). The control culture was a commercial S. thermophilus strain, which was testedalone (CTRL) or in combination with the L. brevis DSM 32386 (CTRL-DSM). The pHevolution, microbiological counts, MiSeq Illumina and UHPLC-HQOMS analysis on milkand cheese samples were performed after 2, 9, and 20 days ripening. During thewhole ripening, the pH was always under 5.5 in all batches. The concentration ofGABA increased during ripening, with the highest content in 84C after 9 days ripening(84 ± 37 mg/kg), in 84C-DSM and CTRL-DSM after 20 days ripening (91 ± 28and 88 ± 24 mg/kg, respectively). The data obtained support the hypothesis thatS. thermophilus 84C and L. brevis DSM 32386 could be exploited as functional cultures,improving the in situ bio-synthesis of GABA during cheese ripening.

Keywords: lactic acid bacteria, model cheese, GABA-enriched cheese, health-promoting bacteria, MiSeqIllumina, Ultra High Performance Liquid Cromatography - Orbitrap Q-Exactive Mass Spectrometry

INTRODUCTION

Milk and dairy products are a good food source of high-quality proteins, minerals and vitamins.Because of the presence of saturated fatty acids, some people believe that dairy foods maybe detrimental to health, and limit or exclude dairy foods from their diet, especially if theyare overweight or predisposed to cardiovascular disease (Rozenberg et al., 2016). However,observational evidences do not support the hypothesis that dairy fat contributes to obesity,

Frontiers in Microbiology | www.frontiersin.org 1 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 2

Carafa et al. Naturally GABA-Enriched Cheese

and its relation with the increase of low-density lipoproteincholesterol and development of cardio-vascular disease is stillunclear (Muehlhoff et al., 2013).

GABA is a non-protein amino acid acting as inhibitoryneurotransmitter in the mammalian central nervous system.GABA has proven effects on brain function, preventing oralleviating anxiety, depression, sleeplessness, memory loss; itstimulates the immune system, prevents inflammation processes,hypertension and diabetes, and regulates the energy metabolism(Dhakal et al., 2012). GABA is naturally present in smallquantities in many vegetal foods, and at high concentration infermented products, especially fermented dairy products andsoy sauces (Diana et al., 2014). The concentration of GABAdetected in 22 different Italian cheese varieties ranged between0.260 and 391 mg/kg (Siragusa et al., 2007), and between 320and 6773.5 mg/kg in Cheddar cheese (Wang et al., 2010; Pouliot-Mathieu et al., 2013). Several authors reported the ability ofselected lactic acid bacteria (LAB) and bifidobacteria to producethis GABA in vitro from the precursor L-glutamic acid (Siragusaet al., 2007; Li and Cao, 2010; Carafa et al., 2015; Franciosiet al., 2015), and investigated the GABA-producing ability ofLAB belonging to S. thermophilus, L. plantarum, L. paracasei,L. delbrueckii subsp. bulgaricus, and Lactococcus lactis infermented cows’ milk and yogurt. The GABA concentrationdetected in the latter studies ranged between 15 and 5000 mg/kg,even though L-glutamate was added to milk before starting thefermentation process (Siragusa et al., 2007; Lacroix et al., 2013;Nejati et al., 2013; Linares et al., 2016). i.e., we reported theability of Lactobacillus brevis BT66 (hereafter DSM 32386) andStreptococcus thermophilus 84C isolated from traditional alpinecheeses, to produce high concentration of GABA (Carafa et al.,2015; Franciosi et al., 2015). In the present study, the hypothesisthat both strains (S. thermophilus 84C as starter and L. brevisDSM 32386 as non-starter) are able to produce GABA in cheeseand to increase the concentration of GABA over ripening (2, 9,and 20 days) was tested. The use of raw milk was chosen forenhancing the natural production of free amino acids (includingL-glutamate) by the proteolytic activity of milk resident bacteriaon the peptides released by the hydrolytic action of the calf renneton caseins (McSweeney, 2004).

To our knowledge, this is the first research addressed to theproduction of naturally GABA-enriched raw milk cheese, wherethe effect of two GABA-producing strains and the ripening timeare considered as GABA producing factors.

MATERIALS AND METHODS

Microorganisms and InoculumPreparationThe strains S. thermophilus 84C and L. brevis DSM 32386belong to the culture collection of the Department of FoodQuality and Nutrition - Fondazione Edmund Mach (FEM,San Michele all’Adige, TN, Italy). Both strains were isolatedfrom traditional alpine cheeses (Nostrano and TraditionalMountain Malga cheese, respectively) and were phenotypically,genotypically, and technologically characterized in previous

studies (Carafa et al., 2015; Franciosi et al., 2015). Both strainsshowed the ability to produce GABA in vitro, which was thereason for interest in the present study.

The commercial S. thermophilus was isolated and purifiedfrom the commercial lyophilized starter mix culture LYOFASTMOT 086 EE (Sacco, Cadorago, CO, Italy), which did notshow any GABA-producing activity in vitro, and was selectedas control strain. The S. thermophilus 84C and the commercialS. thermophilus were grown in M17 broth at 45◦C, and L. brevisDSM 32386 was grown at 30◦C in de Man, Rogosa and Sharpe(MRS) broth, acidified to pH 5.5 with lactic acid 5 mol/L. After48 h incubation, the strains were individually sub-cultured inM17 or MRS (1%, v/v) and grown for 24 h at their optimaltemperature. Cultures were harvested by centrifugation at 3,220 gfor 10 min at 4◦C, the pellets were suspended in Skim Milk(SM, Oxoid; 10:1, v/v; e.g., 400 mL pure culture in 40 mL SM),divided in 1 mL aliquots and frozen in liquid nitrogen. Thecellular concentration of each culture was calculated in triplicateby plate counting onto M17 agar for S. thermophilus strainsand MRS for L. brevis. All media were purchased from Oxoid(Milan, Italy).

Experimental Cheese ManufactureWe produced four mini-cheese batches, as follows: the control(CTRL) mini-cheese was manufactured inoculating raw cow’smilk with the commercial S. thermophilus starter strain; the 84Cmini-cheese (84C) was produced inoculating S. thermophilus84C as starter strain; the CTRL-DSM mini-cheese was producedadding the commercial S. thermophilus as starter strain andL. brevis DSM 32386 as adjunct culture, and the 84C-DSM mini-cheese was produced inoculating S. thermophilus 84C as starterand L. brevis DSM 32386 as adjunct culture. Both commercialand 84C S. thermophilus strains were inoculated into bulk milkat concentration 106 CFU/mL, and L. brevis DSM 32386 atconcentration 101 CFU/mL. Experimental cheeses were producedaccording to the method described by Cipolat-Gotet et al. (2013),with some modifications (Figure 1). Raw cow’s milk (1.5 L)was heated at 37◦C, inoculated with the bacterial mixture andcoagulated with the addition of calf rennet (Naturen Plus 215Hansen, Pacovis Amrein AG, Bern, Switzerland, 215 IMCU/mL).After coagulation, curd was cut into nut-size grains, cooked at48◦C for 10 min, and finally rested at this temperature for 20more min. After separation, draining and molding, curd waspressed for 30 min at room temperature and salted for 30 minin a saturated brine solution (17◦Be, 9◦SH/50, 28◦C). After twodays ripening at 15◦C and 85% relative humidity, all batches werestored under-vacuum and ripened for 18 more days. For eachbatch were produced three wheels to be opened and analysedafter 2, 9 and 20 days ripening, and three replicates for eachtime point were produced in three consecutive weeks for a totalof 36 mini-cheese batches. The chemical composition analysisof milk was performed by MilkoScanTM FT6000 (Foss ElectricA/S, Hillerød, Danimarca), while the chemical composition ofwhey and cheese was determined by the FoodScanTM Lab (FossElectric A/S).

1https://docs.tibco.com

Frontiers in Microbiology | www.frontiersin.org 2 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 3

Carafa et al. Naturally GABA-Enriched Cheese

FIGURE 1 | Description of the model cheese-making process.

Microbiological AnalysisMilk, curd, and cheese samples were submitted to microbiologicalanalysis. Four grams of cheese were homogenized with 36 gof sterile Na-citrate 2% (w/w) solution by ULTRA-TURRAX R©

(IKA R© Werke GmbH & Co., KG, Staufen, Germany) for 5 minat 15,000 rpm, inside the microbiological cabinet. Then, milk andcheese samples were decimally diluted and plated onto selectiveagar media and incubated as follows: MRS agar acidified to pH5.5 with 5 mol/L lactic acid, anaerobiosis, 48 h at 30◦C and 45◦C

for mesophilic and thermophilic LAB rod-shaped, respectively;M17 agar for 48 h, aerobiosis at 30◦C and anaerobiosis at 45◦Cfor mesophilic and thermophilic LAB cocci-shaped, respectively;violet red bile agar (VRBA) for 24 h, aerobiosis at 37◦C forcoliforms; plate count agar (PCA) with SM (10 g/L, w/v) for 24 h,aerobiosis at 30◦C for total bacterial count (TBC). All culturemedia were purchased from Oxoid.

DNA Extraction and MiSeq LibraryPreparationIllumina analysis was performed on all milk and cheesesamples. Ten mL of milk and homogenized cheese samples werecentrifuged at 3,220 g for 15 min at +4◦C. The genomic DNAwas extracted from the pellet using the Power FoodTM MicrobialDNA Isolation Kit (Mo Bio Laboratories Inc., Carlsbad, CA,United States) according to the manufacturer’s instructions. AllDNA samples were purified by PowerClean DNA Clean-up Kit(Mo Bio Laboratories Inc.) and quantified by Nanodrop8800Fluorospectrometer (Thermo Scientific, United States).

A 464-nucleotide sequence of the bacterial V3-V4 region(Baker et al., 2003; Claesson et al., 2010) of the 16S rRNA gene(Escherichia coli positions 341 to 805) was amplified. Uniquebarcodes were attached before the forward primers to facilitatethe pooling and subsequent differentiation of samples. To preventpreferential sequencing of smallest amplicons, the ampliconswere cleaned using the Agencourt AMPure kit (Beckmancoulter) according to manufacturer’s instructions. The DNAconcentration of amplicons was determined using the Quant-iTPicoGreen dsDNA kit (Invitrogen) following the manufacturer’sinstructions. In order to ensure the absence of primer dimersand to assay the purity, the generated amplicon libraries qualitywas evaluated by a Bioanalyzer 2100 (Agilent, Palo Alto, CA,United States) using the High Sensitivity DNA Kit (Agilent).Following quantitation, the cleaned amplicons were mixed andcombined in equimolar ratios. Pair-end sequencing was carriedout at CIBIO (Center of Integrative Biology) – University ofTrento (Trento, Italy) using the Illumina MiSeq system (Illumina,United States).

Illumina Data Analysis and SequencesIdentification by QIIME2Raw paired-end FASTQ files were demultiplexed using idemp2

and imported into Quantitative Insights Into Microbial Ecology(Qiime2, version 2018.2). Sequences were quality filtered,trimmed, de-noised, and merged using DADA2 (Callahan et al.,2016). Chimeric sequences were identified and removed via theconsensus method in DADA2. Representative sequences werealigned with MAFFT and used for phylogenetic reconstructionin FastTree using plugins alignment and phylogeny (Price et al.,2009; Katoh and Standley, 2013). Alpha and beta diversitymetrics were calculated using the core-diversity plugin withinQIIME2 and emperor (Vazquez-Baeza et al., 2013). Taxonomicand compositional analyses were conducted by using pluginsfeature-classifier3. A pre-trained Naive Bayes classifier based on

2https://github.com/yhwu/idemp/blob/master/idemp.cpp3https://github.com/qiime2/q2-feature-classifier

Frontiers in Microbiology | www.frontiersin.org 3 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 4

Carafa et al. Naturally GABA-Enriched Cheese

the Greengenes 13_8 99% Operational Taxonomic Units (OTUs)database4, which had been previously trimmed to the V4 regionof 16S rDNA, bound by the 341F/805R primer pair, was appliedto paired-end sequence reads to generate taxonomy tables.

The data generated by Illumina sequencing were deposited inthe NCBI Sequence Read Archive (SRA) and are available underAc. PRJNA497423.

Quantification of GABA and GlutamateThe amino acid composition of milk and all cheese samples wasquantified by Ultra High Performance Liquid Cromatography- Orbitrap Q-Exactive Mass Spectrometry (UHPLC-HQOMS;UHPLC Ultimate 3000RS, ThermoScientific, Rodano, Italy), atthe Technology Transfer Centre, Fondazione Edmund Mach(FEM, San Michele all’Adige, Italy).

Two grams of cheese were mixed to 0.4 g of sulfosalicylic acid,suspended in 29.7 mL of perchloric acid (0.01 M) and 0.3 mL ofβ-glutamic acid (500 mg/L) and homogenized with a ULTRA-TURRAX R© for 10 min at 15,000 rpm. The suspensions weresubmitted to sonication for 30 min and centrifuged at 3,220 g for20 min. The supernatant was filtered through a 0.22 µm pore sizefilter (Minisart, Sartorius Stedim Biotech, Goettingen, Germany)and diluted 1:50 with a water/methanol solution (50:50, v/v). Formilk, 20 grams of sample were mixed to 0.4 g of sulfosalicylicacid, cooled in ice for 10 min and centrifuged at 3,220 g for10 min. Two grams of supernatant were suspended in 29.7 mLof perchloric acid (0.01 M) and 0.3 mL of β-glutamic acid(500 mg/L) and homogenized. Samples were filtered through a0.22 µm filter (Minisart, Goettingen) and diluted 1:50 with awater/methanol solution (50:50, v/v).

The separation was carried out with formic acid 0.1% (v/v;eluent A; Sigma) and methanol with formic acid 0.1% (v/v; eluentB; Sigma), injecting 5 µL of sample through an Acclaim TrinityP1 column (3 µm particle size, 100 mm × 2.1 mm I.D.; Merk,Germany) at 35◦C. The flow rate was set at 0.4 mL/min. Theanalytical gradient for eluent B was: 1 min at 2%, 4 min at 30%,up to 50% in 0.5 min, to 100% in 0.5 min, held at 100% for 3 minfor cleaning, and to 2% for reconditioning in 0.5 min.

Mass spectra were acquired in positive mode through a fullMS analysis at mass resolving power 70,000. For ionization, HESIII parameters were set as follow: heated capillary temperature to330◦C; sheath gas flow rate at 40 arbitrary units; auxiliary gas flowrate at 20 arbitrary units; spray voltage at 3.0 kV; auxiliary gasheater temperature at 300◦C.

GABA, glutamic acid and the internal standard β-glutamicacid were detected in the extracted ion chromatograms (EICs)corresponding to the protonated molecules [M+H]+ (masstolerance < 5 ppm) used for quantification, whereas dd-MS/MSspectra compared with those collected from available standardswere used for confirmation.

Calibration curves were obtained by plotting the peak arearatio of the quantifier ions (Astandard/Ainternalstandard), multipliedby the internal standard concentration, versus the correspondingconcentration level. Precision, expressed as Relative StandardDeviation (RSD%) of repeatability was tested injecting the same

4http://greengenes.secondgenome.com/

sample four times within the sequence and must be ≤20%(SANTE guidelines 11945, 2015).

Statistical AnalysisOne-way analysis of variance (one-way ANOVA) was performedon all data using STATISTICA data analysis software system,version 13 (TIBCO Software Inc., 2017). Multiple comparison ofmeans was performed using Tukey’s test at a p value of<0.05.

RESULTS

Physico-Chemical Characteristics andMicrobial Counts of the ExperimentalCheeses During the Ripening TimeThe physico-chemical characteristics of milk, whey and cheeseafter 20 days of ripening are shown in Table 1. The differentbatches did not show any significant difference in terms ofchemical composition (p> 0.05). The pH evolution was recordedin milk and during the cheese making process, as shown inFigure 2. The pH value of milk ranged between 6.47 and 6.51and slightly decreased in curd after extraction with no significantdifferences among the four batches (6.41–6.47, p > 0.05).After 2 days ripening a further reduction of pH was observedin all batches, and cheese samples manufactured with thecommercial S. thermophilus culture showed a significantly lowerpH (5.05) than the cheeses produced with the cultures including

TABLE 1 | Chemical composition of milk, whey and cheese used in theexperimental cheese production.

MV SD

Milk

Fat, % 3.58 0.08

Protein, % 3.24 0.02

Lactose, % 5.10 0.01

Caseins, % 2.64 0.02

Total solids, % 12.49 0.09

pH 6.50 0.02

SCS1, units 3.33 0.31

Whey

Fat, % 0.66 0.07

Protein, % 0.87 0.02

Lactose, % 5.08 0.02

pH 6.46 0.03

Cheese

Weight, g 119.18 16.29

Fat, % 32.70 4.49

Protein, % 25.71 2.58

Salt, % 2.04 0.07

Total solids, % 69.73 9.94

pH 5.30 0.32

Moisture, % 29.92 5.95

FDM2, % 47.21 2.30

1SCS: somatic cell count = 3+log2(SCC/100). 2FDM: fat dry matter.

Frontiers in Microbiology | www.frontiersin.org 4 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 5

Carafa et al. Naturally GABA-Enriched Cheese

FIGURE 2 | pH dynamic of the experimental cheeses during ripening. Values are expressed as mean value ± standard deviation. Different letters (a–c) indicate asignificant difference (p < 0.05) among the batches at the same sampling point (Tuckey’s test) indicate significant differences.

S. thermophilus 84C (5.24). The lowest pH value was reached after9 days ripening, with no significant differences (p> 0.05) amongthe four batches (the pH ranged between 4.98 and 5.13). At theend of the ripening period, the pH slightly increased to 5.16–5.34 with no significant differences (p > 0.05) among the cheesebatches studied.

The total bacterial counts in milk, curd, and experimentalcheeses after 2, 9, and 20 days ripening are shown in Table 2.

The count of all microbial groups detected in milk increasedsignificantly (p ≤ 0.001) in the curd with the exception ofcoliforms. The highest total bacterial counts were observedafter 2 days ripening (8.6–9.0 log CFU/g), with no significantdifferences among the different batches. After 9 and 20 daysripening, total bacterial counts showed a little decreasing trendwith no significant differences (p< 0.05) among batches.

LAB (M17 and MRS counts) showed a similar trend of totalaerobic bacteria throughout the ripening. LAB counts increasedfor 2 days, and then decreased approximately one logarithmiccycle (p> 0.05) until the end of ripening.

The load on M17 30◦C showed a trend similar to TBC,without any significant difference among batches. The bacterialconcentration on M17 45◦C was very low in bulk milk (<1log CFU/mL) and increased in curd after inoculating the testedstrains, ranging between 7.2 ± 0.2 and 7.4 ± 0.3 log CFU/g.After 2 days ripening, the count on M17 45◦C increasedby about one logarithmic unit (p > 0.05) in cheese samplesinoculated with the commercial S. thermophilus, and was stablein samples containing the autochthonous S. thermophilus 84C.Milk contained 4.1 ± 0.27 log CFU/mL bacteria on MRS, whichincreased by about 3 and 4–5 logarithmic units in curd andcheese, respectively.

Very low counts on VRBA were detected in milk (1.5± 0.6 logCFU/mL), which reached the maximum count in cheese after 9days ripening (range between 4.7± 0.3 and 5.3± 1.1 log CFU/g).

MiSeq Sequencing DataThe sequences obtained by MiSeq Illumina analysis were firstsubmitted to merging and quality trimming, and 1,741,006reads were subsequently analyzed. After alignment, theremaining Operational Taxonomic Units (OTUs) wereclustered at 3% distance. In order to analyze the bacterialcommunity richness in milk and cheese, the number ofOTUs and the diversity Shannon index were determinedusing QIIME 2 at 97% similarity levels. Data are shownin Table 3.

The differences between samples were also evaluated by BrayCurtis phylogenetic metric (Figure 3). Cheeses inoculated withthe 84C strain (84C and 84C-DSM) clustered together (orangeand violet icons) on the third axis and were separated fromcheeses inoculated with the commercial S. thermophilus starterstrain. The PERMANOVA analysis, showed a higher dissimilarityamong samples inoculated with different S. thermophilus strains(p < 0.01) than samples inoculated with or without L. brevisDSM 32386.

Microbial Communities Identified in Milkand Cheese SamplesThe milk community was mainly constituted byMoraxellaceae (35%), Flavobacteriales (18%) andother Gammaproteobacteria (Chryseobacterium andPseudomonas, 16%) (Figure 4). Streptococcaceae andLactobacillaceae were detected at 8 and 2% relativeabundance, respectively.

The cheese microbiota after 2 days ripening (Figure 4) showedthat Streptococcaceae was the most abundant family in all batches(49% relative abundance in CTRL, 51% in CTRL-DSM, 54% in84C and 56% in 84C-DSM). Furthermore, we observed otherdominant microbial families as Enterococcaceae (29% in CTRL,

Frontiers in Microbiology | www.frontiersin.org 5 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 6

Carafa et al. Naturally GABA-Enriched Cheese

TABLE 2 | Changes in microbial counts of milk and cheese batches throughout ripening (curd, 2, 9, and 20 days).

TBC M17 30◦C M17 45◦C MRS 30◦C VRBA

MV SD MV SD MV SD MV SD MV SD

Milk 5.4a 0.8 5.0a 0.5 0.4a 1.0 4.1a 0.3 1.5a 0.6

CTRL

curd 7.6b 0.9 7.1b 0.3 7.2b 0.2 7.2b 0.3 1.6a 0.9

2 days 8.9c 0.1 8.7c 0.2 8.3b,c 0.2 9.0c 0.2 4.9b,c 1.5

9 days 8.1b,c 0.3 9.0c 1.1 8.0b,c 0.1 8.2b,c 0.4 4.7b 0.6

20 days 8.9c 0.8 7.7b,c 1.2 7.1b 0.7 8.3b,c 1.6 4.4b 0.9

84C

curd 7.6b 0.6 7.2b 0.3 7.4b 0.2 7.3b 0.3 2.0a 1.1

2 days 8.6c 0.4 8.6c 0.4 7.9b,c 0.1 8.9c 0.3 4.4b 0.4

9 days 7.9b 0.3 7.7b,c 0.1 7.1b 0.4 8.0b,c 0.2 5.3c 1.1

20 days 8.3b,c 0.9 7.8b,c 0.3 7.4b 0.3 8.5c 0.9 5.1b,c 0.4

CTRL-DSM

curd 7.3b 0.4 7.3b 0.5 7.2b 0.1 7.1b 0.3 1.9a 1.0

2 days 9.0c 0.4 8.8c 0.4 8.7c 0.3 9.1c 0.4 4.5b 1.6

9 days 8.4b,c 0.2 8.4b,c 0.3 7.7b,c 0.4 8.2b,c 0.4 5.0b,c 0.6

20 days 8.0b,c 0.7 7.8b,c 0.7 7.6b,c 0.2 8.3b,c 1.6 4.9b,c 0.5

84C-DSM

curd 7.8b,c 1.1 7.4b 0.9 7.4b 0.3 7.2b 0.3 2.2a 0.8

2 days 8.7c 0.9 8.6c 0.8 7.6b,c 0.4 8.4b,c 0.8 4.9b,c 1.1

9 days 8.5b,c 0.2 8.5c 0.2 7.9b,c 0.4 8.4b,c 0.5 5.1c 0.5

20 days 8.6c 0.5 7.6b,c 0.4 7.3b 0.3 8.5c 1.4 4.9b,c 0.7

The microbial counts in milk were performed before Streptococcus thermophilus (both commercial and 84C) and Lactobacillus brevis DSM 32386. Control (CTRL) wasproduced adding the commercial S. thermophilus, 84C cheese (84C) was produced inoculating S. thermophilus 84C; DSM cheese (CTRL-DSM) was produced addingthe commercial S. thermophilus and L. brevis DSM 32386, and 84C-DSM cheese (84C-DSM) was produced inoculating S. thermophilus 84C and L. brevis DSM 32386.a, b, c, and d: Different letters in the same column indicate significant statistical differences (Tukey’s Test p < 0.05).

TABLE 3 | Number of sequences analyzed (N reads), diversity richness (Chao 1),Observed OTUs (OTUs), and diversity index (Shannon) for experimental cheesesand milk samples.

Sample N reads Observed OTUs Chao1 Shannon

Milk 17.167 82 84.05 4.76

CTRL

2 days 50.419 235 272.97 6.43

9 days 47.698 193 209.70 5.99

20 days 37.841 155 190.83 5.41

84C

2 days 52.458 288 331.32 6.58

9 days 46.364 311 362.34 6.25

20 days 47.323 261 313.69 6.24

CTRL-DSM

2 days 51.551 261 296.26 6.49

9 days 46.493 229 260.42 6.45

20 days 48.336 239 279.26 6.10

84C-DSM

2 days 47.072 219 248.95 5.75

9 days 44.886 237 275.37 5.94

20 days 42.727 196 226.01 5.47

27% in CTRL-DSM, 15% in 84C and 11% in 84C-DSM) andEnterobacteriaceae (18% in CTRL, 27% in CTRL-DSM, 23% in84C and 25% in 84C-DSM).

After 9 days ripening the microbial composition didnot show significant change, with the exception of cheesesamples inoculated with L. brevis DSM 32386 strain whereLactobacillaceae were detected at 1.9 (CTRL-DSM) and 0.9%(84C-DSM), while in the other cheeses this microbial family wasunder 0.01% relative abundance.

At the end of ripening there was a change of the microbiota,in particular in cheese samples inoculated with the commercialS. thermophilus starter. We observed a significant decreasein Streptococcaceae from 45 to 30% and an increase ofEnterococcaceae from 27 to 47% that became the dominantmicrobial family in this cheese. Enterobacteriaceae, rangingbetween 11 and 14%, decreased in all batches (Figure 4).

GABA Detection in Milk and CheeseSamplesThe GABA concentrations are shown in Figure 5A. The UHPLC-HQOMS analysis on milk and cheese samples showed an increaseof GABA during ripening, with the highest content in 84C after9 days ripening (84 ± 37 mg/kg), in 84C-DSM and CTRL-DSMafter 20 days ripening (91± 28 and 88± 24 mg/kg, respectively).Milk contained 1.9 ± 0.9 mg/kg GABA, which slightly increased(p > 0.05) in all cheese batches after 2 days ripening. The firstsignificant change was observed in CTRL-DSM, 84C-DSM and84C samples after 9 days ripening, which contained significantlymore GABA (63± 19, 47± 22, and 84± 37 mg/kg, respectively)

Frontiers in Microbiology | www.frontiersin.org 6 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 7

Carafa et al. Naturally GABA-Enriched Cheese

FIGURE 3 | Bray Curtis analysis of milk and cheese samples after 2, 9, and 20 days ripening. Red dots show milk samples (n = 3), the blue dots show the CTRLcheese (n = 9), green dots show CTRL-DSM cheese (n = 9), purple dots show 84C cheese (n = 9), and orange dots show 84C-DSM cheese samples (n = 9).

than CTRL (9.1 ± 7.8 mg/kg, p < 0.05). At the end of ripening(20 days), 84C-DSM cheese had the highest level of GABA(91 ± 22 mg/Kg), which was not significantly different fromCTRL-DSM (88 ± 20 mg/Kg) and 84C (73 ± 21 mg/Kg), butsignificantly higher than CTRL cheese (11± 10 mg/Kg).

Glutamic acid was also detected in milk (36 ± 7 mg/kg)and cheese during ripening (Figure 5B). After 2 days ripeningthe content of glutamate ranged between 36 ± 10 and62 ± 39 mg/kg, and increased after 9 days ripening between47 ± 17 and 117 ± 8 mg/kg. We observed a significantlyhigher content of glutamate in 84C-DSM after 9 days ripening(117 ± 8 mg/kg), compared to CTRL (64 ± 37 mg/kg) and84C batches (47 ± 17 mg/kg). At the end of ripening, theconcentration of glutamate fluctuated between 48 ± 38 and114 ± 40 mg/kg glutamate, with no significant differencesbetween batches.

DISCUSSION

In the present study, the use of raw milk was chosen in order toexploit the ability of the resident microbiota to release naturallyfree amino acids (including L-glutamate) from the peptidesoriginated from the hydrolytic action of the calf rennet oncaseins (McSweeney, 2004).

The experimental mini-cheese batches were produced in thelaboratory, developing a cheese making protocol which couldsimulate as better the technology of alpine raw milk cheesesin very small vats. The processing of small volumes of milk(1.5 L) has the advantage that up to 20 cheese batches/day canbe produced using the same milk, but the disadvantage that thewheels are very small, and the high surface/volume ratio makesthe cheese subjected to a fast drying process. For this reasonthe ripening process was conducted under-vacuum in order tomimic the anaerobic conditions of the cheese core and avoidingthe formation of thick rinds.

All cheese batches had similar chemical composition at the endof ripening, suggesting that the tested strains operated similarly.

The pH evolution was monitored during the cheese makingprocess because the role of pH is not only related to the outcomeof the fermentation, but also to the production of GABA. Thecommercial S. thermophilus reduced the pH of cheese in ashorter time than the strain S. thermophilus 84C within 2 daysripening. Conversely, any of the tested strain (both alone andin combination with L. brevis DSM 32386) had not statisticallysignificant effect (p> 0.05) on pH after 9 and 20 days ripening.

With regards to the microbiological concentration, weobserved an increase of the plate counts in curd samples for allmicrobial batches, which is probably due to the physical retentionof bacteria in the curd and to the multiplication of the inoculated

Frontiers in Microbiology | www.frontiersin.org 7 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 8

Carafa et al. Naturally GABA-Enriched Cheese

FIGURE 4 | Relative abundance of the microbial taxonomic groups detected in milk and cheese samples after 2, 9, and 20 days ripening. Data are expressed inpercentage.

strains during the coagulation phase. The addition of L. brevisDSM 32386 was not detectable on MRS 30◦C, and it might berelated to the initial concentration of lactobacilli in raw milk(4.1 log CFU/mL), which was similar to the concentration of theinoculated L. brevis.

Analyzing the sequences obtained by MiSeq Illumina, weobtained much information looking at the quality of the readsand alpha and beta-diversity. The rarefaction analysis based onOTUs at 97% similarity showed approximation to asymptote, andwe concluded that the recovered sequences represented properlythe diversity of the bacterial communities in the 39 samples.

Based on the OTUs number and the Shannon and Chaoindexes, there was no difference related to the inoculated strain.There was a decreasing richness and diversity of microbialpopulation with fermentation time. In fact, the number ofOTUs and Chao and Shannon indexes were higher in samplescollected after 2 than after 20 days but these differences were notsignificant. The presence or absence of L. brevis DSM 32386 didnot significant influence the Bray Curtis index, as also confirmedby PERMANOVA analysis.

Both alpha- and beta-diversity analysis suggested that L. brevishad no effect on the microbiota of experimental cheesesat the concentration used in this work. Furthermore, thedifferences in cheese microbiota are likely due by the starter

used even if belonging to the same species and added at thesame concentration.

The presence in milk of Chryseobacterium and Pseudomonaswhich are gram-negative spoilage bacteria was not unexpectedbecause they are usually present in raw milk and areable to grow at low temperature during refrigerated storage(Quigley et al., 2013; von Neubeck et al., 2015). Conversely,Streptococcaceae and Lactobacillaceae, which are involved indairy fermentation and in the determination of organoleptic,flavor and texture properties of the final product (Quigley et al.,2013) are totally desired. The dominance of Streptococcaceaeafter 2 days ripening suggested that bacteria belonging tothis family (including the added S. thermophilus strains)started the fermentation process. After 9 days ripeningLactobacillaceae were significantly higher in CTRL-DSM and84C-DSM cheeses, suggesting that the sequences identified asLactobacillaceae corresponded to the inoculated L. brevis DSM32386. Taxonomy data at the end of ripening suggested thatonly the commercial S. thermophilus starter strain influencedsignificantly the cheese microbiota and that all the testedstrains had a decreasing effect on the relative abundanceof Enterobacteriaceae.

During cheese ripening, L-glutamate which is naturallypresent in caseins (Zoon and Allersma, 1996) might be

Frontiers in Microbiology | www.frontiersin.org 8 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 9

Carafa et al. Naturally GABA-Enriched Cheese

FIGURE 5 | GABA (A) and L-glutamic acid (B) concentration detected in milk and cheese samples after 2, 9, and 20 days ripening. The mean value ± standarddeviation of both amino acids is expressed in mg/kg. Different letters (a, b, c) indicate a significant difference (p < 0.05) among the batches at the same samplingpoint (Tuckey’s test) indicate significant differences.

released from caseins proteolysis. In cheese, the ripeningperiod can facilitate this process, and L-glutamate can beconverted to GABA by GABA-producing bacteria. For thisreason we chose cheese as GABA-carrier food, even thoughmany factors play a key role on the final content of GABA,like as the cheese-making process, the presence of starteror adjunct cultures and the ripening conditions (Siragusaet al., 2007). In order to determine when GABA is producedand if it persist during ripening, we decided to monitor thecontent of GABA at different ripening stages. We detectedvery interesting variations of GABA and glutamate betweenthe four batches. High production of GABA by fermentationis correlated to the activity of glutamic acid decarboxylase(GAD) but also on the inhibition of GABA-decomposingenzymes. GABA transaminase (GABA-T) promotes thereversible conversion of GABA to succinic semi-aldehyde using

either pyruvate-dependent GABA transaminase (GABA-TP)or α-ketoglutarate-dependent GABA transaminase (GABA-TK), and succinic semi-aldehyde dehydrogenase catalyzes thereversible conversion of succinic semi-aldehyde to succinate(Takayama and Ezura, 2015).

Overlapping GABA and glutamate concentration wehypothesized that the presence of L. brevis DSM 32386 enhancedthe release of glutamate from caseins between day 2 and day 20,and the consequent conversion of L-glutamate to GABA. In fact,the increase of glutamate in CTRL-DSM samples after 9 and20 days ripening corresponded to the increase of GABA. Theseresults suggested that glutamate was gradually consumed over itsrelease and converted to GABA.

By contrast, S. thermophilus 84C showed a different trend;the GABA production significantly (p < 0.05) increased in 84Ccheese samples between 2 and 9 days ripening, but it was related

Frontiers in Microbiology | www.frontiersin.org 9 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 10

Carafa et al. Naturally GABA-Enriched Cheese

to a slight reduction of L-glutamate. Afterwards, between day9 and day 20, we observed a slight decrease of GABA and anincrease of L-glutamate 84C samples. These data suggest thatthe strain S. thermophilus 84C is likely not involved in therelease of caseins, and that GABA is metabolized to succinateand reconverted to L-glutamate after reentering the Krebs cycle.This hypothesis is supported by 84C-DSM cheese samples, wherewe observed a constant increase of GABA over all the ripeningperiod and a strong and significant decrease (p < 0.05) ofL-glutamate between 9 and 20 days ripening. Since we did notperform any metatranscriptome or gene expression analysis, weare not able to confirm this hypothesis, and more investigationneeds to be done.

Several studies demonstrated that GABA can reduce highblood pressure in animals and humans, as reviewed by Dianaet al. (2014). In human intervention trials, a 12 weekstreatment with 100 ml of fermented milk containing between10 and 12 mg of GABA, or 50 g of GABA-enriched cheesecontaining 16 mg of GABA, decreased blood pressure inhypertensive patients (Inoue et al., 2003; Pouliot-Mathieuet al., 2013). The experimental cheese produced in thepresent study contained about 117 mg/kg, thus a portionof 100 g would provide about 11.7 mg of GABA, coveringthe total intake needed in order to detect positive effects onhuman health.

The strains tested in the present study would be useful forthe production of bioactive traditional alpine cheeses, whichare mostly produced from raw milk. Cheese-makers fromTrentino Alps are strongly faithful to tradition, and they are notwilling to make any change that could compromise or modifythe biodiversity and sensory characteristics of their cheeses.Whereas, they usually agree when they are suggested to useautochthonous strains that are able to improve the product.In this contest, a naturally GABA enriched cheese producedfrom raw alpine milk would satisfy the needs of both producersand consumers, who increasingly ask for healthy foods, andthe marketing of local dairy products would rise. On the otherhand, the utilization of raw milk might affect the productionof GABA in the experimental cheeses, through a synergisticeffect between the indigenous microbiota and the tested GABA-producing strains. From an industrial point of view, the use ofpasteurized milk would facilitate the optimization of the GABAproduction in cheese, the standardization and reproducibilityof results. For this reason, a new project aiming to develop

a naturally GABA-enriched cheese from pasteurized milk isin progress.

CONCLUSION

The aim of this study was to demonstrate that GABA isproduced in raw milk cheese and accumulates during ripening.The data obtained showed that after 20 days ripening, 84C-DSM, CTRL-DSM and 84C cheeses had a GABA contentsignificantly higher than CTRL cheese, confirming the hypothesisthat S. thermophilus 84C and L. brevis DSM 32386 could beexploited as bio-functional cultures, facilitating the in situ bio-synthesis of GABA during cheese ripening, and providing anoption to replace chemical GABA with natural GABA.

We need to take into account that the production ofexperimental mini-cheeses had some limitations, which are thesize of the wheels and the ripening conditions. The ripening tookplace in aerobic condition only during the first two days ripeningand from day 3 to day 20 under-vacuum. The small size allowedproducing many cheese batches within the same cheese-makingprocess, while the storage and ripening under-vacuum had thedouble purpose of miming the anaerobic conditions of the cheesecore and avoiding the total drying of cheese. However, this is thefirst study addressed to the manufacture of GABA-enriched rawmilk cheese by LAB and more trials will be carried out in orderto optimize the production and the accumulation of GABA incheese and reduce as much as possible its degradation.

DATA AVAILABILITY

Most of the relevant data are included within the manuscript.Any additional raw data supporting the conclusions will be madeavailable by the authors on request, without undue reservation, toany qualified researcher.

AUTHOR CONTRIBUTIONS

EF and GB devised the study. IC and EF drafted the manuscript.EF, IC, GS, and TN performed the experiments. EF, GB, RL,and KT provided resources and intellectual input that supportedthe study.

REFERENCESBaker, G. C., Smith, J. J., and Cowan, D. A. (2003). Review and re-analysis of

domain-specific 16S primers. J. Microbiol. Methods 55, 541–555. doi: 10.1016/j.mimet.2003.08.009

Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., andHolmes, S. P. (2016). DADA2: high-resolution sample inference from Illuminaamplicon data. Nat. Methods 13, 581–583. doi: 10.1038/nmeth.3869

Carafa, I., Nardin, T., Larcher, R., Viola, R., Tuohy, K., and Franciosi, E. (2015).Identification and characterization of wild lactobacilli and pediococci fromspontaneously fermented Mountain cheese. Food Microbiol. 48, 123–132. doi:10.1016/j.fm.2014.12.003

Cipolat-Gotet, C., Cecchinato, A., De Marchi, M., and Bittante, G. (2013). Factorsaffecting variation of different measures of cheese yield and milk nutrientrecovery from an individual model cheese-manufacturing process. J. Dairy Sci.96, 7952–7965. doi: 10.3168/jds.2012-6516

Claesson, M. J., Wang, Q., O’sullivan, O., Greene-Diniz, R., Cole, J. R., Ross,R. P., et al. (2010). Comparison of two next-generation sequencing technologiesfor resolving highly complex microbiota composition using tandem variable16S rRNA gene regions. Nucleic Acids Res. 38:e200. doi: 10.1093/nar/gkq873

Dhakal, R., Bajpai, V. K., and Baek, K. H. (2012). Production of GABA(γ-aminobutyric acid) by microorganisms: a review. Braz. J. Microbiol. 43,1230–1241. doi: 10.1590/S1517-83822012000400001

Frontiers in Microbiology | www.frontiersin.org 10 February 2019 | Volume 10 | Article 93

fmicb-10-00093 February 12, 2019 Time: 16:37 # 11

Carafa et al. Naturally GABA-Enriched Cheese

Diana, M., Quílez, J., and Rafecas, M. (2014). Gamma-aminobutyric acid as abioactive compound in foods: a review. J. Funct. Foods 10, 407–420. doi: 10.1016/j.jff.2014.07.004

Franciosi, E., Carafa, I., Nardin, T., Schiavon, S., Poznanski, E., Cavazza, A.,et al. (2015). Biodiversity and γ-aminobutyric acid production by lactic acidbacteria isolated from traditional alpine raw cow’s milk cheeses. BioMed. Res.Int. 2015:625740. doi: 10.1155/2015/625740

Inoue, K., Shirai, T., Ochiai, H., Kasao, M., Hayakawa, K., Kimura, M., et al.(2003). Blood-pressure-lowering effect of a novel fermented milk containinggamma-aminobutyric acid (GABA) in mild hypertensives. Eur. J. Clin. Nutr.57, 490–495. doi: 10.1038/sj.ejcn.1601555

Katoh, K., and Standley, D. M. (2013). MAFFT multiple sequence alignmentsoftware version 7: improvements in performance and usability Mol. Biol. Evol.30, 772–780. doi: 10.1093/molbev/mst010

Lacroix, N., St-Gelais, D., Champagne, C. P., and Vuillemard, J.-C. (2013). Gamma-aminobutyric acid-producing abilities of lactococcal strains isolated from old-style cheese starters. Dairy Sci. Technol. 93, 315–327. doi: 10.1007/s13594-013-0127-4

Li, H., and Cao, Y. (2010). Lactic acid bacterial cell factories for gamma-aminobutyric acid. Amino Acids 39, 1107–1116. doi: 10.1007/s00726-010-0582-7

Linares, D. M., O’Callaghan, T. F., O’Connor, P. M., Ross, R. P., andStanton, C. (2016). Streptococcus thermophilus APC151 strain is suitable for themanufacture of naturally GABA-enriched bioactive yogurt. Front. Microbiol.7:1876. doi: 10.3389/fmicb.2016.01876

McSweeney, P. L. (2004). Biochemistry of cheese ripening. Int. J. Dairy Technol. 57,127–144. doi: 10.1111/j.1471-0307.2004.00147.x

Muehlhoff, E., Bennett, A., and McMahon, D. (2013). Milk and Dairy Productsin Human Nutrition. Rome: Food and Agriculture Organization of the UnitedNations.

Nejati, F., Rizzello, C. G., Di Cagno, R., Sheikh-Zeinoddin, M., Diviccaro, A.,Minervini, F., et al. (2013). Manufacture of a functional fermented milkenriched of Angiotensin-I Converting Enzyme (ACE)-inhibitory peptides andγ-amino butyric acid (GABA). Food Sci. Technol. 51, 183–189. doi: 10.1016/j.lwt.2012.09.017

Pouliot-Mathieu, K., Gardner-Fortier, C., Lemieux, S., St-Gelais, D., Champagne,C. P., and Vuillemard, J.-C. (2013). Effect of cheese containing gamma-aminobutyric acid-producing acid lactic bacteria on blood pressure in men.Pharmanutrition 1, 1–8. doi: 10.1016/j.phanu.2013.06.003

Price, M. N., Dehal, P. S., and Arkin, A. P. (2009). FastTree: computing largeminimum evolution trees with profiles instead of a distance matrix. Mol. Biol.Evol. 26, 1641–1650. doi: 10.1093/molbev/msp077

Quigley, L., O’sullivan, O., Stanton, C., Beresford, T. P., Ross, R. P., Fitzgerald, G. F.,et al. (2013). The complex microbiota of raw milk. FEMS Microbiol. Rev. 37,664–698. doi: 10.1111/1574-6976.12030

Rozenberg, S., Body, J. J., Bruyere, O., Bergmann, P., Brandi, M. L., Cooper, C.,et al. (2016). Effects of dairy products consumption on health: benefitsand beliefs—a commentary from the belgian bone club and the europeansociety for clinical and economic aspects of osteoporosis, osteoarthritis andmusculoskeletal diseases. Calcif. Tissue Int. 98, 1–17. doi: 10.1007/s00223-015-0062-x

SANTE guidelines 11945 (2015), Available at: http://services.accredia.it/extsearch_documentazione.jsp?area55&ID_LINK707&page12&IDCTX5067&id_context5067

Siragusa, S., De Angelis, M., Di Cagno, R., Rizzello, C. G., Coda, R., andGobbetti, M. (2007). Synthesis of γ-aminobutyric acid by lactic acid bacteriaisolated from a variety of Italian cheeses. Appl. Environ. Microbiol. 73, 7283–7290. doi: 10.1128/AEM.01064-07

Takayama, M., and Ezura, H. (2015). How and why does tomato accumulate a largeamount of GABA in the fruit? Front. Plant Sci. 6:612. doi: 10.3389/fpls.2015.00612

TIBCO Software Inc. (2017). Statistica (Data Analysis Software System), Version 13.Availabel at: http://statistica.io

Vazquez-Baeza, Y., Pirrung, M., Gonzalez, A., and Knight, R. (2013). EMPeror:a tool for visualizing high-throughput microbial community data. Gigascience2:16. doi: 10.1186/2047-217X-2-16

von Neubeck, M., Baur, C., Krewinkel, M., Stoeckel, M., Kranz, B., Stressler, T., et al.(2015). Biodiversity of refrigerated raw milk microbiota and their enzymaticspoilage potential. Int. J. FoodMicrobiol. 211, 57–65. doi: 10.1016/j.ijfoodmicro.2015.07.001

Wang, H. K., Dong, C., Chen, Y. F., Cui, L. M., and Zhang, H. P.(2010). A new probiotic cheddar cheese with high ACE-inhibitoryactivity and γ-aminobutyric acid content produced with koumiss-derived Lactobacillus casei Zhang. Food Technol. Biotechnol. 48,62–70.

Zoon, P., and Allersma, D. (1996). Eye and crack formation in cheese bycarbon dioxide from decarboxylation of glutamic acid. Neth. Milk Dairy J. 50,309–318.

Conflict of Interest Statement: The authors declare that the research wasconducted in the absence of any commercial or financial relationships that couldbe construed as a potential conflict of interest.

Copyright © 2019 Carafa, Stocco, Nardin, Larcher, Bittante, Tuohy and Franciosi.This is an open-access article distributed under the terms of the Creative CommonsAttribution License (CC BY). The use, distribution or reproduction in other forumsis permitted, provided the original author(s) and the copyright owner(s) are creditedand that the original publication in this journal is cited, in accordance with acceptedacademic practice. No use, distribution or reproduction is permitted which does notcomply with these terms.

Frontiers in Microbiology | www.frontiersin.org 11 February 2019 | Volume 10 | Article 93